Metrics_ How to Improve Key Business Results - Martin Klubeck [69]
I agree with this concept and would happily argue that you can have higher confidence in any metric made up of multiple measures than one made up of less. Answers derived from other metrics encourage an even higher level of confidence.
I don't propose that we seek out this level of confidence by demanding meta-metrics (metrics made up of other metrics) at every turn. Remember, some questions may not dictate even a single metric. Some may simply need a yes or no answer. Others may need a measure (or two) to provide the necessary insight to make a better decision. But, the more complex and critical the question, the more likely a higher level of confidence is called for. If your root question will be used to help decide on the future direction of the entire organization along with a considerable investment, you might demand a higher level of confidence in individual data used to build the metric and in the metric's overall conclusions.
Conflicting Results
Because we use varying methods and data sources, we run the real risk of obtaining conflicting results. But, rather than see this as a negative, you should see it as a positive.
Let's look at our restaurant example. If our restaurant's effectiveness metric is made up of Delivery, Usage, and Customer Satisfaction, we may expect that the results of each of these measurement areas should always coincide. If we have good service (Delivery) we should have high Customer Satisfaction ratings. And if we have high Customer Satisfaction, we could assume that we should have high levels of repeat customers, and high usage. We also expect the opposite. If we have poor Customer Satisfaction, we expect that customers won't come back. If we don't deliver well (too slow, wrong items delivered, or the menu items are unavailable) we would expect poor ratings and less usage.
These are logical assumptions, but many times incorrect ones. Each of the permutations tells us something different. In Table 7-2, let's look at each measure with a simple high vs. low result. Of course, the real results of your measures may be much more complicated—especially when you remember that each can also utilize triangulation. Delivery could have high availability and speed to deliver, but poor accuracy. You could have high usage for one type of clientele and low for another. Customer satisfaction could have high marks for some areas (courtesy of staff) and low for others (efficiency of staff). Rather than complicate it further, let's look at the measures at the higher level, keeping in mind the complexity possible when taking into account lower levels of triangulation.
Besides the interpretation of each of the individual measurement areas, the triangulation itself offers information you would lack if you only collected one or two areas.
Using our restaurant business as an example, let's interpret some possible measures. In Table 7.2, you can see how different permutations of the results of the measures can tell a different story. While each measure provides some basic insights, it is more meaningful to look at them in relation to each other.
You may argue that triangulation seems to make the results more confusing, not clearer. But in actuality, triangulation assures that you have more data and more views of that data. The more information you have the better your answers will be. But in all cases the next step should be the same. Investigate, investigate, investigate. The beauty of triangulation is that you already have so many inputs that your investigation can be much more focused and reap greater benefits with less additional work.
Imagine if all you measured was Customer Satisfaction. If you ratings in this area were high, what could you determine? You could think life was good. But if you're not making enough money to keep your business open, you'll wonder what happened.
Triangulation